Malawi

library(sf)
library(readxl)
library(dplyr)
library(plyr)
library(ggplot2)
library(afrihealthsites)
library(ggpubr)
library(afriadmin)
library(tmap)
library(cowplot)
library(forcats)
library(colorspace)

# Install Malawi MFL
malawi_MFL = read_excel("~/malawi-health-facilities-1/MHFR_Facilities 1.xlsx")

# Convert to sf

## omit NA's
new_malawi_MFL = na.omit(malawi_MFL)

## check for NA 
any(is.na(new_malawi_MFL))

## [1] FALSE

## transform geometry columns into numeric 
sapply(new_malawi_MFL, class)

##        CODE        NAME COMMON NAME   OWNERSHIP        TYPE      STATUS 
## "character" "character" "character" "character" "character" "character" 
##        ZONE    DISTRICT DATE OPENED    LATITUDE   LONGITUDE 
## "character" "character" "character" "character" "character"

new_malawi_MFL = transform(new_malawi_MFL, LATITUDE = as.numeric(LATITUDE), 
                                           LONGITUDE = as.numeric(LONGITUDE))

## Warning in eval(substitute(list(...)), `_data`, parent.frame()): NAs introduced
## by coercion

any(is.na(new_malawi_MFL)) ## check for NA

## [1] TRUE

new_malawi_MFL = na.omit(new_malawi_MFL) ## and omit

## convert to sf object
malawi_facilities_MFL = st_as_sf(new_malawi_MFL, coords = c("LONGITUDE", "LATITUDE"), dim = "XY")

malawi_facilities_MFL = st_set_crs(malawi_facilities_MFL, 4326) ## set CRS, is WGS84 right?

head(malawi_facilities_MFL)

## Simple feature collection with 6 features and 9 fields
## geometry type:  POINT
## dimension:      XY
## bbox:           xmin: 33.74129 ymin: -15.84 xmax: 35.09 ymax: -13.79742
## geographic CRS: WGS 84
##       CODE                                               NAME  COMMON.NAME
## 1 MC010002                               A + A private clinic          A+A
## 2 BT240003                                       A-C Opticals A.C Opticals
## 3 BT240005                                Akwezeke PVT Clinic Akwezeke Pvt
## 4 BT240006                                  AB Medical Clinic        Abowa
## 5 LL040007                                 ABC Comm. Hospital   ABC Clinic
## 6 LL040010 Achikondi Women Community Friendly Services Clinic    Achikondi
##                                       OWNERSHIP       TYPE     STATUS
## 1                                       Private     Clinic Functional
## 2                                       Private     Clinic Functional
## 3                                       Private     Clinic Functional
## 4                                       Private     Clinic Functional
## 5 Christian Health Association of Malawi (CHAM)   Hospital Functional
## 6                                       Private Dispensary Functional
##                 ZONE DISTRICT DATE.OPENED                   geometry
## 1 Centrals West Zone  Mchinji  Jan 1st 75 POINT (33.88563 -13.79742)
## 2    South East Zone Blantyre  Jan 1st 75        POINT (35.03 -15.8)
## 3    South East Zone Blantyre  Jan 1st 75       POINT (35.09 -15.84)
## 4    South East Zone Blantyre  Jan 1st 75       POINT (35.09 -15.84)
## 5 Centrals West Zone Lilongwe  Jan 1st 75 POINT (33.74129 -13.96816)
## 6 Centrals West Zone Lilongwe  Jan 1st 75  POINT (33.7793 -13.95473)

Overview:

Malawi MFL

Facility locations managed by the Ministry of Health and Population Malawi. WHO guidance states that it should be updated at least every 2 years. Last update is unknown.
Locations in lat-long, CRS set to WGS84
1546 facilities reported, data includes name, ownership, type of facility and functional status.
1426 facilities after removing NAs
9 types of facilities, including unclassified. Number of each type is shown. Top 3 are clinics (585), health centers (520) and Dispensaries (179). All are primary level. What is private?

# Re-order the facility types
malawi_MFL$TYPE = as.factor(malawi_MFL$TYPE)

malawi_MFL$TYPE = factor(malawi_MFL$TYPE, levels = c("Central Hospital", "District Hospital", "Hospital", "Health Centre", "Clinic", "Health Post", "Dispensary", "Private", "Unclassified"))

# Number of each type + ownership
facility_types_MFL = as.data.frame(table(malawi_MFL$TYPE, malawi_MFL$OWNERSHIP))
head(facility_types_MFL)

##                Var1            Var2 Freq
## 1  Central Hospital Aquaid Lifeline    0
## 2 District Hospital Aquaid Lifeline    0
## 3          Hospital Aquaid Lifeline    0
## 4     Health Centre Aquaid Lifeline    0
## 5            Clinic Aquaid Lifeline    0
## 6       Health Post Aquaid Lifeline    0

levels(facility_types_MFL$Var2)

## [1] "Aquaid Lifeline"                              
## [2] "Christian Health Association of Malawi (CHAM)"
## [3] "Government"                                   
## [4] "Mission/Faith-based (other than CHAM)"        
## [5] "Non-Government"                               
## [6] "Other"                                        
## [7] "Parastatal"                                   
## [8] "Private"

# Only showing government, private and CHAM ownership
facility_types_MFL$new_Var2 = revalue(facility_types_MFL$Var2, c("Aquaid Lifeline"="Other", "Non-Government"="Other", "Parastatal"="Other", "Mission/Faith-based (other than CHAM)"="Other"))
facility_types_MFL$new_Var2 = factor(facility_types_MFL$new_Var2, levels = c("Government", "Private", "Christian Health Association of Malawi (CHAM)", "Other"))


## bar plot of no. of facility types 
plot_facility_types_MFL = ggplot(facility_types_MFL, aes(x=Var1, y=Freq, fill=new_Var2)) + geom_bar(position = "stack", stat = "identity")

plot_facility_types_MFL = plot_facility_types_MFL + labs(x = "Facility types", y = "Frequency", fill="Ownership") + scale_fill_brewer(palette = "Set2") + coord_flip() + theme_minimal() + ggtitle("MFL") + theme(axis.title.y = element_blank(), legend.title = element_text(size = 8, face = "bold"), legend.text = element_text(size = 8), plot.title = element_text(face = "bold", hjust = 0), legend.position = "bottom")
plot_facility_types_MFL

par(mar=c(11,4,4,4))

8 types of ownership, top 3 are government (695), Private (495) and CHAM. It seems that private not for profit is separated from private for profit in this data set.

# Re-order ownership
malawi_MFL$OWNERSHIP = as.factor(malawi_MFL$OWNERSHIP)

malawi_MFL$OWNERSHIP = factor(malawi_MFL$OWNERSHIP, levels = c("Government", "Private", "Christian Health Association of Malawi (CHAM)", "Non-Government", "Mission/Faith-based (other than CHAM)", "Other", "Parastatal", "Aquaid Lifeline"))

# Number of each type of ownership
ownership_MFL = as.data.frame(table(malawi_MFL$OWNERSHIP))

## bar plot of ownership
plot_ownership_MFL = ggplot(ownership_MFL, aes(x=forcats::fct_relabel(Var1,stringr::str_wrap,width = 16), y=Freq)) + geom_bar(stat = "identity", fill="slategray")

plot_ownership_MFL = plot_ownership_MFL + labs(x="Ownership", y = "Frequency") + coord_flip() + theme_minimal() + ggtitle("MFL") + theme(axis.title.y = element_blank(), plot.title = element_text(face = "bold"))
plot_ownership_MFL

WHO-KWTRP

Focuses on facilities run by government, faith-based organisations, NGO’s and local authorities. Covers 50 countries in sub-Saharan Africa. Sources of information include health sector reports, websites run by national or international organisations and personal communications
If MFL was available it was used. More than one datasource was often used to compile facility list
Private facilities are excluded, duplicates removed, name errors corrected and name variations were matched. Missing info was added with the use of other datasources.
Now hosted by the WHO Global Malaria Programme, last update February 2019
Malawi datasources includes MFL, https://data.humdata.org/dataset/malawi-health and http://www.cham.org.mw/uploads/7/3/0/8/73088105/cham_health_facilities_-_1_june_2016.pdf
At time of publishing, 639 facilities with 9 missing coordinates, not been updated since
Data includes facility name, type, ownership, source of location and reclassified facility types

# Malawi WHO data.frame
malawi_WHO <- afrihealthsites("malawi", datasource='who', plot=FALSE, returnclass='dataframe')

head(malawi_WHO)

## # A tibble: 6 x 10
##   Country Admin1 `Facility name` `Facility type` Ownership   Lat  Long
##   <chr>   <chr>  <chr>           <chr>           <chr>     <dbl> <dbl>
## 1 Malawi  Centr… 80 Block Clinic Clinic          MoH       -12.9  33.4
## 2 Malawi  Centr… ABC Community … Clinic          FBO       -14.0  33.7
## 3 Malawi  Centr… Adventist Heal… Health Centre   FBO       -14.0  33.8
## 4 Malawi  Centr… Alinafe Commun… Community Hosp… FBO       -13.4  34.2
## 5 Malawi  Centr… Area 18 Health… Health Centre   MoH       -13.9  33.8
## 6 Malawi  Centr… Area 25 Health… Health Centre   MoH       -13.9  33.8
## # … with 3 more variables: `LL source` <chr>, iso3c <chr>,
## #   facility_type_9 <chr>

Original facility types: 8 types, top 3 are health centers (457), health post/dispensary (87) and mission hospitals (27). Large difference between types.

# Re order facility types and ownership 
malawi_WHO$`Facility type` = as.factor(malawi_WHO$`Facility type`)
malawi_WHO$`Facility type` = factor(malawi_WHO$`Facility type`, levels = c("Central Hospital", "District Hospital", "Mission Hospital", "Rural Hospital", "Community Hospital", "Health Centre", "Clinic", "Health Post/Dispensary"))

malawi_WHO$Ownership = as.factor(malawi_WHO$Ownership)
malawi_WHO$Ownership = factor(malawi_WHO$Ownership, levels = c("MoH", "FBO", "Local authority", "NGO"))

# No. of original facility types + ownership
facility_types_WHO = as.data.frame(table(malawi_WHO$`Facility type`, malawi_WHO$Ownership))
head(facility_types_WHO)

##                 Var1 Var2 Freq
## 1   Central Hospital  MoH    4
## 2  District Hospital  MoH   24
## 3   Mission Hospital  MoH    0
## 4     Rural Hospital  MoH   17
## 5 Community Hospital  MoH    0
## 6      Health Centre  MoH  332

## bar plot of original facility types 
plot_facility_types_WHO = ggplot(facility_types_WHO, aes(x=Var1, y=Freq, fill=Var2)) + geom_bar(position = "stack", stat = "identity")
plot_facility_types_WHO = plot_facility_types_WHO + labs(x = "Facility types", y = "Frequency", fill="Ownership") + scale_fill_brewer(palette = "Set2") + coord_flip() + theme_minimal() + ggtitle("WHO") + theme(axis.title.y = element_blank(), legend.title = element_text(size = 8, face = "bold"), legend.text = element_text(size = 8), plot.title = element_text(face = "bold", hjust = 0), legend.position = "bottom") + expand_limits(y=c(0,600))
plot_facility_types_WHO

Reclassified facility types: Hospitals are aggregated except Community Hospital (became Community Health Unit), the rest are the same.

# Re order
malawi_WHO$facility_type_9 = as.factor(malawi_WHO$facility_type_9)
malawi_WHO$facility_type_9 = factor(malawi_WHO$facility_type_9, levels = c("Hospital", "Health Centre", "Health Clinic", "Health Post", "Community Health Unit"))

# No. of reclassified facility types
RC_facility_types_WHO = as.data.frame(table(malawi_WHO$facility_type_9))
RC_facility_types_WHO

##                    Var1 Freq
## 1              Hospital   80
## 2         Health Centre  457
## 3         Health Clinic   22
## 4           Health Post   87
## 5 Community Health Unit    2

## bar plot of reclassified facility types 
plot_RC_facility_types_WHO = ggplot(RC_facility_types_WHO, aes(x=Var1, y=Freq)) + geom_bar(stat = "identity", fill="slategray")
plot_RC_facility_types_WHO = plot_RC_facility_types_WHO + labs(x = "Reclassified facility types", y = "Frequency") + coord_flip() + theme_minimal() + theme(axis.title.y = element_blank(), plot.title = element_text(face = "bold")) + ggtitle("WHO") + expand_limits(y=c(0,600))
plot_RC_facility_types_WHO

Ownership: 4 types, majority is government owned (467) with faith based organisations second.

# Types of ownership
ownership_WHO = as.data.frame(table(malawi_WHO$Ownership))
ownership_WHO

##              Var1 Freq
## 1             MoH  467
## 2             FBO  173
## 3 Local authority    5
## 4             NGO    3

## bar plot of ownership
plot_ownership_WHO = ggplot(ownership_WHO, aes(x=Var1, y=Freq)) + geom_bar(stat = "identity", fill="slategray")
plot_ownership_WHO = plot_ownership_WHO + labs(x="Ownership", y = "Frequency") + coord_flip() + theme_minimal() + ggtitle("WHO") + theme(axis.title.y = element_blank(), plot.title = element_text(face = "bold")) + expand_limits(y=c(0,600))
plot_ownership_WHO

Both data sources contain no information on services available, capacity or equipment. MFL does state whether facility is functional.

Classification of MFL facilities aligns more with the structure of the health care system in Malawi (community, primary, secondary, tertiary), it differentiates central hospitals from district and other hospitals. WHO has additional rural and mission hospitals, where do they fit in?

https://www.health.gov.mw/index.php/2016-01-06-19-58-23/national-aids states that at community level, health posts, dispensaries and maternity clinics offer services. Primary includes health centers and community hospitals, secondary consists of district and some CHAM hospitals, tertiary includes central hospitals.

Analysis:

1546 compared to 648 facilities, difference of 898

Ownership

Maps

## tmap mode set to interactive viewing

## Warning in sf::st_is_longlat(shp2): bounding box has potentially an invalid
## value range for longlat data

## Warning in sf::st_is_longlat(shp2): bounding box has potentially an invalid
## value range for longlat data

## Warning in sf::st_is_longlat(shp2): bounding box has potentially an invalid
## value range for longlat data

Qs to address?:

How many facilities are in the same location across the MFL and WHO datasets?

2 points intersect directly and so are same up to 5 decimal places (WHO data has all coordinates to 5 decimal places). Check fewer decimal places to see if more intersect?

Do same facilities share same names/other attributes?

Ownership is the same for both facilities but Euthini is registered as a rural hospital in the MFL data while in the WHO data, it is a health centre. The name reflects this.

# Qs 1 - how many intersect?

## convert malawi_WHO to sf object

class(malawi_WHO)

## [1] "tbl_df"     "tbl"        "data.frame"

any(is.na(malawi_WHO))

## [1] TRUE

new_malawi_WHO = na.omit(malawi_WHO) ## omit NA

sf_malawi_WHO = st_as_sf(new_malawi_WHO, coords = c("Long", "Lat"), dim = "XY")
sf_malawi_WHO = st_set_crs(sf_malawi_WHO, 4326)

## st_intersection
intersect_WHO_MFL = st_intersection(x=sf_malawi_WHO, y=malawi_facilities_MFL)

## although coordinates are longitude/latitude, st_intersection assumes that they are planar

## Warning: attribute variables are assumed to be spatially constant throughout all
## geometries

intersect_WHO_MFL ## only 2 intersect directly, so are same up to 5 decimal places?

## Simple feature collection with 2 features and 17 fields
## geometry type:  POINT
## dimension:      XY
## bbox:           xmin: 33.29456 ymin: -11.53894 xmax: 33.41925 ymax: -11.45836
## geographic CRS: WGS 84
## # A tibble: 2 x 18
##   Country Admin1 Facility.name Facility.type Ownership LL.source iso3c
## * <chr>   <chr>  <chr>         <fct>         <fct>     <chr>     <chr>
## 1 Malawi  North… Euthini Heal… Health Centre MoH       GPS       MWI  
## 2 Malawi  North… Madede Healt… Health Centre MoH       GPS       MWI  
## # … with 11 more variables: facility_type_9 <fct>, CODE <chr>, NAME <chr>,
## #   COMMON.NAME <chr>, OWNERSHIP <chr>, TYPE <fct>, STATUS <chr>, ZONE <chr>,
## #   DISTRICT <chr>, DATE.OPENED <chr>, geometry <POINT [°]>

The ones that aren’t, are they within 50m of another facility?

# Try merge_points()
merge_points("malawi", datasources=c('who', 'healthsites'), dist_same_m=50)

## although coordinates are longitude/latitude, st_nearest_points assumes that they are planar

## malawi num points: who : 639 healthsites : 266  shared at dist thresh 50 m : 52  after merging: 853

How many facilities per admin1 and admin2 regions?

Top 3 districts for MFL are: Lilongwe (163 facilities), Blantyre (154) and Mzimba (95)
Top 3 districts for WHO are: Mzimba (56 facilities), Lilongwe (52) and Mangochi (45)

# Facilities per region 

# MFL, admin1
facility_admin1_MFL = st_intersects(malawi_admin1, malawi_facilities_MFL, sparse = TRUE)
malawi_admin1$facility = lengths(facility_admin1_MFL)
head(malawi_admin1) # gave no. of facilities per region

## Simple feature collection with 3 features and 6 fields
## geometry type:  MULTIPOLYGON
## dimension:      XY
## bbox:           xmin: 32.67036 ymin: -17.12952 xmax: 35.91857 ymax: -9.368326
## geographic CRS: WGS 84
##         shapeName shapeISO           shapeID shapeGroup shapeType
## 1  Central Region     MW-C MWI-ADM1-3_0_0-B1        MWI      ADM1
## 2 Southern Region     MW-S MWI-ADM1-3_0_0-B2        MWI      ADM1
## 3 Northern Region     MW-N MWI-ADM1-3_0_0-B3        MWI      ADM1
##                         geometry facility
## 1 MULTIPOLYGON (((33.5486 -12...      508
## 2 MULTIPOLYGON (((34.58346 -1...      609
## 3 MULTIPOLYGON (((34.02092 -1...      248

# Intersect opposite to get admin1 region for each facility 
admin1_facility_MFL = unlist(st_intersects(malawi_facilities_MFL, malawi_admin1, sparse = TRUE))

# error when adding admin1_facility_MFL data to malawi_facilities_MFL, 
# no. of rows are less in admin1_, why?


# trying to figure out which points don't intersect 

intersect_admin1_MFL = st_intersection(malawi_facilities_MFL, malawi_admin1) # returns data frame of matches

intersect_admin1_MFL = intersect_admin1_MFL[-c(10:16)] # remove columns 

# adding region data for admin1 to intersect_admin1_MFL (using admin1_facility_MFL from previously)
intersect_admin1_MFL$region_admin1 = admin1_facility_MFL

# working out which facilities did not intersect
intersect_admin1_MFL = as.data.frame(intersect_admin1_MFL) # convert to data frame to use in anti_join
head(intersect_admin1_MFL)

##       CODE                                               NAME
## 1 MC010002                               A + A private clinic
## 5 LL040007                                 ABC Comm. Hospital
## 6 LL040010 Achikondi Women Community Friendly Services Clinic
## 7 LL040011                        Lilongwe Adventist Hospital
## 8 LL040012                    Adventist Health Centre Area 15
## 9 LL040013                         Africa Leaf Clinic Kanengo
##                  COMMON.NAME                                     OWNERSHIP
## 1                        A+A                                       Private
## 5                 ABC Clinic Christian Health Association of Malawi (CHAM)
## 6                  Achikondi                                       Private
## 7    Adventist Health Centre Christian Health Association of Malawi (CHAM)
## 8                  Adventist Christian Health Association of Malawi (CHAM)
## 9 Africa Leaf Clinic Kanengo                                       Private
##            TYPE         STATUS               ZONE DISTRICT DATE.OPENED
## 1        Clinic     Functional Centrals West Zone  Mchinji  Jan 1st 75
## 5      Hospital     Functional Centrals West Zone Lilongwe  Jan 1st 75
## 6    Dispensary     Functional Centrals West Zone Lilongwe  Jan 1st 75
## 7      Hospital     Functional Centrals West Zone Lilongwe  Jan 1st 83
## 8 Health Centre     Functional Centrals West Zone Lilongwe  Jan 1st 75
## 9        Clinic Non-functional Centrals West Zone Lilongwe  Jan 1st 75
##                     geometry region_admin1
## 1 POINT (33.88563 -13.79742)             1
## 5 POINT (33.74129 -13.96816)             2
## 6  POINT (33.7793 -13.95473)             2
## 7  POINT (33.7793 -13.95473)             2
## 8  POINT (33.7793 -13.95473)             1
## 9  POINT (33.80487 -13.8898)             1

no_intersect_admin1_MFL = anti_join(malawi_facilities_MFL, intersect_admin1_MFL) # shows which did not intersect 

# input 0s in the region_admin1 column
no_intersect_admin1_MFL$region_admin1 = rep.int(0, times = 61)

# turn back into sf 
no_intersect_admin1_MFL = st_as_sf(no_intersect_admin1_MFL) 
intersect_admin1_MFL = st_as_sf(intersect_admin1_MFL) 

# bind data frames together to get new data frame with admin1 regions assigned correctly
malawi_facilities_MFL_admin1 = rbind(intersect_admin1_MFL, no_intersect_admin1_MFL)
tail(malawi_facilities_MFL_admin1)

## Simple feature collection with 6 features and 10 fields
## geometry type:  POINT
## dimension:      XY
## bbox:           xmin: 1 ymin: -33.45 xmax: 26.56 ymax: -1
## geographic CRS: WGS 84
##          CODE                      NAME               COMMON.NAME
## 5610 MN281537 GOOD SAMARITAN PVT CLINIC GOOD SAMARITAN PVT CLINIC
## 5710 MC011538        Kazyzyo Dispensary       Kazyozyo Dispensary
## 5810 MH201539              Cape Maclear                    Chembe
## 5910 MH201540                 Nankhwali                 Nankhwali
## 6010 MH201541      Kapire Health Centre                    Kapire
## 6110 ZA231548          Prime PVT CLINIC                  Jali pvt
##                                          OWNERSHIP          TYPE
## 5610                                       Private       Private
## 5710                                    Government    Dispensary
## 5810                                    Government   Health Post
## 5910 Christian Health Association of Malawi (CHAM) Health Centre
## 6010 Christian Health Association of Malawi (CHAM) Health Centre
## 6110                                       Private       Private
##                                      STATUS               ZONE DISTRICT
## 5610                             Functional    South East Zone   Mwanza
## 5710                             Functional Centrals West Zone  Mchinji
## 5810 Pending Operation (Under construction)    South West Zone Mangochi
## 5910                             Functional    South West Zone Mangochi
## 6010                             Functional    South West Zone Mangochi
## 6110                             Functional    South West Zone    Zomba
##      DATE.OPENED region_admin1             geometry
## 5610  Jan 1st 75             0         POINT (3 -2)
## 5710  Jan 1st 90             0 POINT (26.56 -33.45)
## 5810  Jan 1st 75             0         POINT (2 -1)
## 5910  Jan 1st 75             0         POINT (2 -1)
## 6010  Jan 1st 75             0         POINT (1 -1)
## 6110  Jan 1st 75             0         POINT (1 -1)

# !!! something is wrong, facilities with same district show different 
# value for region. Was data for district inputted incorrectly or are 
# coordinates wrong? Carry on for now.
                    
# use new data frame for plots


# Map with admin1 layer 
tmap_admin1_MFL = tmap_facilities_MFL + tm_shape(st_geometry(malawi_admin1)) + tm_borders()
tmap_admin1_MFL

## Map - number of facilities per region
tmap_admin1_MFL2 = tm_shape(st_geometry(malawi_admin1)) + tm_borders() + tm_shape(malawi_admin1) + tm_fill("facility", style = "cat", palette = sequential_hcl(4, palette = "YlGnBu", rev = TRUE), title = "MFL facilities")

tmap_admin1_MFL2

# Plot of facilities in each admin1 region 

## transform columns
malawi_facilities_MFL_admin1$region_admin1 = as.factor(malawi_facilities_MFL_admin1$region_admin1)
malawi_facilities_MFL_admin1$TYPE = as.factor(malawi_facilities_MFL_admin1$TYPE)

## freq of facility types by region
df_facility_admin1_MFL = as.data.frame(table(malawi_facilities_MFL_admin1$region_admin1, 
                                             malawi_facilities_MFL_admin1$TYPE))
head(df_facility_admin1_MFL)

##   Var1              Var2 Freq
## 1    0  Central Hospital    0
## 2    1  Central Hospital    0
## 3    2  Central Hospital    4
## 4    3  Central Hospital    0
## 5    0 District Hospital    1
## 6    1 District Hospital    7

## subsetting by region for plots 

# region 1 = Central 
admin1_region1 = filter(df_facility_admin1_MFL, Var1 == 1)

admin1_region1_plot = ggplot(admin1_region1, aes(x=Var2, y=Freq)) + geom_bar(stat = "identity", fill="slategray") + labs(x="Facility type", y="Frequency", title = "MFL Central Region") + theme(plot.title = element_text(face = "bold")) + theme_minimal() + coord_flip() + expand_limits(y=c(0, 250))

# region 2 = Southern
admin1_region2 = filter(df_facility_admin1_MFL, Var1 == 2)

admin1_region2_plot = ggplot(admin1_region2, aes(x=Var2, y=Freq)) + geom_bar(stat = "identity", fill="slategray") + labs(x="Facility type", y="Frequency", title = "MFL Southern Region") + theme(plot.title = element_text(face = "bold")) + theme_minimal() + coord_flip() + expand_limits(y=c(0,250))

# region 3 = Northern
admin1_region3 = filter(df_facility_admin1_MFL, Var1 == 3)

admin1_region3_plot = ggplot(admin1_region3, aes(x=Var2, y=Freq)) + geom_bar(stat = "identity", fill="slategray") + labs(x="Facility type", y="Frequency", title = "MFL Northern Region") + theme(plot.title = element_text(face = "bold")) + theme_minimal() + coord_flip() + expand_limits(y=c(0, 150))


# MFL, admin2 
facility_admin2_MFL = st_intersects(malawi_admin2, malawi_facilities_MFL, sparse = TRUE)
malawi_admin2$facility = lengths(facility_admin2_MFL)

sum(malawi_admin2$facility) # also has 1365 intersections

## [1] 1365

# Map - number of facilities per district
tmap_admin2_MFL = tm_shape(st_geometry(malawi_admin2)) + tm_borders() + tm_shape(malawi_admin2) + tm_fill("facility", palette = sequential_hcl(6, palette = "YlGnBu", rev = TRUE), title = "MFL facilities") + tm_layout(title = "MFL")

tmap_admin2_MFL

# WHO, admin1

# Map with admin1 layer
tmap_admin1_WHO = tmap_facilities_WHO + tm_shape(st_geometry(malawi_admin1)) + tm_borders()
tmap_admin1_WHO

## intersecting for Map
facility_admin1_WHO = st_intersects(malawi_admin1, sf_malawi_WHO, sparse = TRUE)
malawi_admin1$facility_WHO = lengths(facility_admin1_WHO)
head(malawi_admin1)

## Simple feature collection with 3 features and 7 fields
## geometry type:  MULTIPOLYGON
## dimension:      XY
## bbox:           xmin: 32.67036 ymin: -17.12952 xmax: 35.91857 ymax: -9.368326
## geographic CRS: WGS 84
##         shapeName shapeISO           shapeID shapeGroup shapeType
## 1  Central Region     MW-C MWI-ADM1-3_0_0-B1        MWI      ADM1
## 2 Southern Region     MW-S MWI-ADM1-3_0_0-B2        MWI      ADM1
## 3 Northern Region     MW-N MWI-ADM1-3_0_0-B3        MWI      ADM1
##                         geometry facility facility_WHO
## 1 MULTIPOLYGON (((33.5486 -12...      508          238
## 2 MULTIPOLYGON (((34.58346 -1...      609          271
## 3 MULTIPOLYGON (((34.02092 -1...      248          130

## Map - number of facilities per region
tmap_admin1_WHO2 = tm_shape(st_geometry(malawi_admin1)) + tm_borders() + tm_shape(malawi_admin1) + tm_fill("facility_WHO", style = "cat", title = "WHO facilities", palette = sequential_hcl(4, palette = "YlGnBu", rev = TRUE))

tmap_admin1_WHO2

# freq of facility types by region
malawi_WHO$Admin1 = as.factor(malawi_WHO$Admin1)
df_facility_admin1_WHO = as.data.frame(table(malawi_WHO$Admin1, 
                                             malawi_WHO$`Facility type`))
head(df_facility_admin1_WHO)

##       Var1              Var2 Freq
## 1  Central  Central Hospital    1
## 2 Northern  Central Hospital    1
## 3 Southern  Central Hospital    2
## 4  Central District Hospital    9
## 5 Northern District Hospital    6
## 6 Southern District Hospital    9

# Central
admin1_central = filter(df_facility_admin1_WHO, Var1 == "Central")

admin1_central_plot = ggplot(admin1_central, aes(x=Var2, y=Freq)) + geom_bar(stat = "identity", fill="slategray") + labs(x="Facility type", y="Frequency", title = "WHO Central Region") + theme(plot.title = element_text(face = "bold")) + theme_minimal() + coord_flip() + expand_limits(y=c(0, 250))

# Southern
admin1_southern = filter(df_facility_admin1_WHO, Var1 == "Southern")

admin1_southern_plot = ggplot(admin1_southern, aes(x=Var2, y=Freq)) + geom_bar(stat = "identity", fill="slategray") + labs(x="Facility type", y="Frequency", title = "WHO Southern Region") + theme(plot.title = element_text(face = "bold")) + theme_minimal() + coord_flip() + expand_limits(y=c(0,250))

# Northern
admin1_northern = filter(df_facility_admin1_WHO, Var1 == "Northern")

admin1_northern_plot = ggplot(admin1_northern, aes(x=Var2, y=Freq)) + geom_bar(stat = "identity", fill="slategray") + labs(x="Facility type", y="Frequency", title = "WHO Northern Region") + theme(plot.title = element_text(face = "bold")) + theme_minimal() + coord_flip() + expand_limits(y=c(0, 150))

# WHO, admin2 
facility_admin2_WHO = st_intersects(malawi_admin2, sf_malawi_WHO, sparse = TRUE)
malawi_admin2$facility_WHO = lengths(facility_admin2_WHO)

# Map - number of facilities per district 
tmap_admin2_WHO = tm_shape(st_geometry(malawi_admin2)) + tm_borders() + tm_shape(malawi_admin2) + tm_fill("facility_WHO", palette = sequential_hcl(4, palette = "YlGnBu", rev = TRUE), title = "WHO facilities")

tmap_admin2_WHO

# final plots
plot_admin1_central = align_plots(admin1_region1_plot, admin1_central_plot, align = "v")
ggdraw(plot_admin1_central[[1]])

ggdraw(plot_admin1_central[[2]])

plot_admin1_southern = align_plots(admin1_region2_plot, admin1_southern_plot, align = "v")
ggdraw(plot_admin1_southern[[1]])

ggdraw(plot_admin1_southern[[2]])

plot_admin1_northern = align_plots(admin1_region3_plot, admin1_northern_plot, align = "v")
ggdraw(plot_admin1_northern[[1]])

ggdraw(plot_admin1_northern[[2]])

Automating the process of sorting facility types into 4 tiers as (Falchetta et al., 2020). Tier 1: dispensary or basic health post; Tier 2: health center; Tier 3: provincial hospital or regional hospital; and Tier 4: central hospital or national hospital.

Function that recognises the different facility types and in different languages (downside = similar to manual process, hard to account for all variations)
Use code/id/reference codes provided with data, do they refer to type of facility? Couldn’t find how the MoH codes with the MFL data for Malawi were assigned.
Use satellite imagery to determine size of building and use that to sort into tiers (downside = many ways this can go wrong, what would the criteria be?)
Use capacity data if available?

# Function that recognises different facility types 
sort_facilities = function(data, facility_column) {
   for(i in levels(facility_column)) {
      if(grepl("Central", facility_column, fixed = TRUE) == TRUE){
         output = grep("Central", facility_column, fixed = TRUE)
         t4 = slice(data, c(output[]))
      }
  return(t4)
 }
}

Malawi

35762384

20/01/2021